Abstract—Ultrasound images are widely used for diagnosis of liver cirrhosis. In liver cirrhosis classification on M-mode ultrasound images, the use of higher order local auto-correlation (HLAC) features has been shown to be effective. In the previous study, we used the traditional 25 dimensional HLAC features. The 25 HLAC features are made by 25 mask patterns with up to 0th, 1st, and 2nd-order. On the other hand, there exists an extension of HLAC features. The extended HLAC features were shown to be more effective when higher-order HLAC features were used. Therefore, by the use of the extended HLAC features, we expected the liver cirrhosis classification performance to improve. However, the effectiveness of the extended HLAC features to classify the liver cirrhosis images is not clear. In this paper, more effectively to classify liver cirrhosis M-mode ultrasound images, we examine the performance of extended HLAC features. Index Terms—Liver cirrhosis classification, M-mode ultrasound images, HLAC features, extended HLAC features. I. INTRODUCTION Ultrasound images are widely used for diagnosis of liver cirrhosis. The liver cirrhosis classification method on M-mode ultrasound images is required. Fig. 1 shows M-mode ultrasound images. Fig. 1(a) is normal. On the other hand, (b) is cirrhosis. In the previous study, Zhou’s method has been proposed [1]. This method consists of 2 processes. Firstly, an abdominal aorta wall from the M-mode ultrasound image is extracted. Secondly, the feature vector generated by based on the extracted abdominal aorta wall is used for liver cirrhosis classification. Hayashi et al. have also proposed a method to extract an abdominal aorta wall accurately [2]. The method of Hayashi et al. is based on a weighted coefficient of correlation of the abdominal aorta wall. It outperforms the Zhou’s method. However, Zhou’s and Hayashi et al. approaches depend on the accuracy of the abdominal aorta wall extraction. If the extraction of the abdominal aorta wall fails, the subsequent liver cirrhosis classification fails by all means. Therefore, we have examined a method to classify the liver cirrhosis not using the abdominal aorta wall extraction process [3], [4]. In the proposed method, we used higher-order local auto-correlation (HLAC) [5] features. The HLAC features are successfully applied to pattern recognition Manuscript received October 9, 2014; revised February 10, 2015. Yoshihiro Mitani is with the Department of Intelligent System Engineering, National Institute of Technology, Ube College, Ube, Japan (e-mail: [email protected]). Yusuke Fujita, Yoshihiko Hamamoto, and Isao Sakaida are with Yamaguchi University, Ube, Japan. problems. The advantages of the HLAC features are considered to be simple, robust, and easily implemented. Furthermore, in order to improve the liver cirrhosis classification performance, we have also proposed to apply image processing techniques [6] of a thresholding technique and a shading method [3], [4]. These methods are expected effectively to reduce noises in the image. In general, the ultrasound images have heavy noises. By the combination of the adaptive thresholding and a shading technique, the HLAC feature vector was effectively extracted [3], [4]. (a) Normal (b) Cirrhosis Fig. 1. M-mode ultrasound images. The dimensionality of the traditional HLAC feature vector is 25. The 25 HLAC feature vector is produced by 25 mask patterns with up to 0th, 1st, and 2nd-order. In the paper [7], the extended HLAC feature vector approach was proposed. The HLAC features are extended by from zeroth-order to eighth-order. This means the dimensionalities of the HLAC features are from 1 to 223. The experimental results showed the effectiveness of the higher-order HLAC features for texture image classification [7]. The higher dimensionality is, the more classification performance improves. That is, it is expected that the more complicated the mask patterns are, the richer the classification information becomes. Therefore, we expected the extended HLAC feature vector approach to further improve the liver cirrhosis classification performance on the M-mode ultrasound images. In this paper, we examine the extended HLAC feature vector approach. Experimental result shows a surprised result. The lower-order HLAC feature shows effective. Thus, in the liver cirrhosis M-mode ultrasound image classification by the use of the HLAC Classification of Liver Cirrhosis on m-Mode Ultrasound Images by Extended Higher Order Local Autocorrelation Features Yoshihiro Mitani, Yusuke Fujita, Yoshihiko Hamamoto, and Isao Sakaida International Journal of Computer Theory and Engineering, Vol. 8, No. 2, April 2016 167 DOI: 10.7763/IJCTE.2016.V8.1038
4
Embed
Classification of Liver Cirrhosis on m-Mode Ultrasound Images … · In liver cirrhosis classification on M-mode ultrasound images, the use of higher order local auto-correlation
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Abstract—Ultrasound images are widely used for diagnosis of
liver cirrhosis. In liver cirrhosis classification on M-mode
ultrasound images, the use of higher order local auto-correlation
(HLAC) features has been shown to be effective. In the previous
study, we used the traditional 25 dimensional HLAC features.
The 25 HLAC features are made by 25 mask patterns with up to
0th, 1st, and 2nd-order. On the other hand, there exists an
extension of HLAC features. The extended HLAC features were
shown to be more effective when higher-order HLAC features
were used. Therefore, by the use of the extended HLAC
features, we expected the liver cirrhosis classification
performance to improve. However, the effectiveness of the
extended HLAC features to classify the liver cirrhosis images is
not clear. In this paper, more effectively to classify liver
cirrhosis M-mode ultrasound images, we examine the
performance of extended HLAC features.
Index Terms—Liver cirrhosis classification, M-mode